Application of a Hybrid CNN-Random Forest Model in the Early Detection and Classification of Parkinson's Disease

In this study, the point where machine learning combines Convolutional Neural Networks (CNN) and Random Forest (RF) models help to identify Parkinson's Disease (PD) in a very early stage. A model of PD diagnosis is designed mainly to improve the accuracy in detecting PD. It involves Convolutional Neural Networks (CNN), which is responsible for extracting image features and physiological data at a high level to add to the diagnosis system. In addition, the feature of RF (Random Forest) in generating a large number of classes to classify effectively is a factor to be considered. The proposed model was tested on five stages of PD, improving accuracy by 92%. The accuracy probability is given from 85%-95%, and the recall probability is 87% to 93 %. The proposed extensive dataset metadata that combined several medical imaging and physiological measures from patients with PD, provide a controlled and precise testing. Model performance is then analyzed by using different measures such as precision, accuracy, recall, and F1 score. Moreover, different visualizing tools, such as confusion matrix and ROC curves were leveraged. The ROC curves produced relatively high areas under the curve with values ranging from 0. 88 to 0. 97 in different phases. This hybrid approach for diagnosing Parkinson's disease (PD) thrives on solving these prevalent issues of dataset imbalance and model overfitting. It employs methods such as SMOTE for data resolution and regularization to enhance the model's generalization capacity. The outcomes suggest using hybrid approaches, including Convolutional Neural Networks (CNN) and Random Forest (RF), to get a good tool for reliable and accurate detection of Parkinson's Disease.